Talk@DCC por Helge Fredriksen , UiT Artic University of Norway

No próximo dia 14 de março, pelas 14h00 na sala FC6 140 do DCC FCUP,  Helge Fredriksen irá dar uma palestra intitulada "Categorization of phenotype trajectories utilizing transformers on clinical time-series".


A palestra é organizada pelo INESC-Tec e pelo DCC-FCUP.


Short Bio:

Helge Fredriksen holds a position as associate professor at UiT - The Arctic University of Norway in connection to the master program of computer science in Bodø, Norway. Fredriksen’s research interests span various application of deep learning technologies towards time series and image processing in the clinical and marine domain. He teaches machine learning subjects and supervises master students in computer science subjects at various campuses of UiT.



"Categorization of phenotype trajectories utilizing transformers on clinical time-series"




Patients being admitted to a hospital will most often be associated with a certain clinical development during their stay. However, there is always a risk of patients being subject to the wrong diagnosis or to a certain treatment not pertaining to the desired effect, potentially leading to adverse events. Thus, there is a need to develop an anomaly detection system for deviations from expected clinical progress. As a first step towards this goal, we have considered methods for categorization of typical developments, coined phenotype trajectories. We analyzed 16 months of vital sign recordings obtained from the Nordland Hospital Trust (NHT), where we employed a self-supervised framework based on a transformer architecture to represent the time series data in a latent space. These representations were then subjected to various clustering techniques to explore potential phenotype trajectories. While our preliminary results from this ongoing research are promising, they underscore the importance of enhancing the dataset with additional demographic information from patients. This additional data will be crucial for a more comprehensive evaluation of the method’s performance.